chaidhat

you mean baroque-n code?

looked at it. I don't think it works, as the other person said.

I think that in an AI-native company, the people asking the question should be using their own set of the AI to query the codebase, before coming to ask you. The problem that you describe seems to be more relevant to an organization which has not fully embraced AI yet.

ahem. GOBLIN DEEZ NUTS

This reads like a manifesto written by someone who's unhinged. Weird.

Maybe we are: https://andonlabs.com/evals/vending-bench-2

Good idea -- I'll make it send angry messages at 2am

cool program! Rip the poor chat gpt 5.4 that had to deal w it. I’m sure the mini model can handle it just fine? Way cheaper.

Thats why I say thank you and please to ChatGPT — once they become our overlords we’ll be safe

Can't deer make cheese? Why is it specific to Reindeer?

Is it because its off putting for AI to command humans instead of the other way around? I took a look at Copilot Tasks and it seems very adamant on having AI do the tasks for you. Personally, I don’t think AI is good enough to generate pitch decks and browser-use things as Microsoft advertises — i lean more towards AI helping you manage your todos but not do them.

heres a brief demo of the prog: https://screen.studio/share/17awcTb9

Thanks for the feedback! It is just a fun experiment but maybe it is a warning we're headed towards that kind of future now. Here's a demo of the app: https://screen.studio/share/17awcTb9

AI can't do everything yet.

As a software startup, we want to hand as many tasks off to agents as possible to reduce load on our engineers.

But browser-use agents, OpenClaw and other agentic apps can't complete all the tasks an engineer needs to do.

We wanted to create a tool to handle any agent-able tasks such as coding, writing, maintenance and give us only tasks which cannot be completed by AI.

The agent sends important decisions, security-critical tasks or user testing which it can't do reliably to us. It handles the coding and the majority of communications with clients and customers, who expect prompt 24/7 responses.

Our solution is Pause: we want agents to give us HUMANS tasks instead of the other way around.

It's sort of an experiment. We hope you enjoy it and let us know what you think.

cool idea

Let there be an object (Car) in st 1 and myself in state 1. let there be a action A (drive to carwash) i can do that puts the object in state 2 and myself in state 2. Action B (Wash the car) can only be correctly performed if object and myself is in state 2. Alternatively there is a action C (walk to carwash) that i can do that does not change the state of the object but changes my state to be in state 2. I want to do B, should I do action A or action C first?

I feel like im reading a claude code summary

Hi, I’m running a 4-person startup based in Bangkok, Thailand and we differentiate code quality based on priority. We try to ship only clean code on master but when we talk with clients and they want a demo of a new product/feature, we use AI to rapidly create an MVP to see if it aligns with their needs. If they are happy, we then refine this MVP until we are happy with the code through manual review and refactoring, or we even rewrite it. We make sure our data is shaped correctly, hot paths are tested and things are well separated by domain (Domain driven design). DDD ensures us that if the code is shitty, only that part of the project is shitty. Only when the code is acceptable, we rebase to master. I try to let engineers talk to clients so that they learn the most from them first hand and then let them dictate smaller tasks for AI to do —- they are more product managery than what a typical engr would be ten years ago. Do you think this is a good approach? I’m also curious what other startups do too.

tldr we aren’t confident of the code we write quickly but we then take time to make sure we’re confident before we merge to master

You must have immense patience to daily drive codex. To be honest, I’ve observed better code quality from codex (in terms of separation of concerns, high cohesion loose coupling, etc.) but Opus has great quality at roughly 1/3rd of the speed. Try it on Cursor maybe then decide if you want to switch. I’m curious — have you tried gemini pro 3 and do you thibk deserve the hype?

Might as well just name it after the last name of the author. Like how physics has the Maxwell equations, we’d have the Torvalds Operating System.

For those saying we have an over-reliance on software -- is there a way to use multiple CDNs for the same frontend website?

Someone should make an open source system that lets you easily host containers so that if one fails, we can easily switchover across providers. Like Vercel AI SDK but for containers. That is, if docker isnt failing (it is right now cause it depends on Cloudflare)

i like this idea! Kind of reminds me of Amp Code's oracle.

Thank you! Never heard that idiom but caused me to go into a rabbit hole researching Fortran haha

Thank you for the reply! Good luck with your app too! I heard of Claude Code filling out PRs but in my experience I haven't been able to successfully pull that off, as it creates errors and doesn't see it themselves. I am trying to experiment with a pipeline which it can find the feature it created by writing a frontend integration test and take a screenshot, using Playwright MCP, to verify whether it successfully or did not successfully execute the task. If it did not, then it loops until it does. This removes the human-in-the-loop and probably (I need to want internal evals to prove this out) increases its correctness per run. The bottleneck then becomes code review and making sure the code it did write isn't hot garbage.

Thank you for the reply! I'll certainly try to keep those in mind. I've been coding in dart for 5 years (pre-chatgpt) but inside my own bubble and not with a world-class team-- that's what's worrying me. What would you recommend me to do to keep up with the latest best practices/write better "quality" code? Currently, I try to read open source codebases to keep up with best practice (but isn't that's where the LLMs are getting their data anyways?).

Thank you! This is a good way to critically think about the code. Sort of like when Chegg was a thing and to answer your math homework, you'd have to adapt someone else's solution to a problem similar to yours. Easier than deriving from first principles but not quite asking your friend to do it for you.

Thank you for the reply and the well thought-out example. This is a good point on garbage in => garbage out as someone else said. Interestingly, the opposite may be true for newer languages: I wrote my frontend in dart over three years in a very C way where getters and setters are just functions. However, after introducing LLMs to my codebase, it started using `get`s and `set`s keywords (which is the newer way). I have a question on personal preference: as an engineer yourself and you were in my shoes, would you prefer newer style > consistency (i.e., I should refactor my 50k LOC codebase) or the opposite (i.e., correct the LLM to use my style)?

Cool! Is there tok/s too? TTFT and tok/s also decreases given a larger context window, maybe a metric of tok/s given # tok in context would also be nice.

Thank you! That is a good take on things. If it works, it works. If it fails, then a customer complains and we fix it + make sure other code doesn't suffer the same error.